[Dovecot] alternate rawlog directory
Hello,
In rc2 sources, the rawlog directory seem to be hard-coded to be looked for in $HOME.
I'd like to have a way to centralized the place where rawlog files are, in order to be able :
. to rotate them more easily . not to leave them at the user's reach . not to have to enter the user's directory to read them
I'd like to have a layout similar to this :
/var/log/dovecot.rawlog/ /user1/20060708-093811-45581.in /user1/20060708-093811-45581.ou /user1/... /user2/20060708-093811-45590.in /user2/20060708-093811-45590.out
What do you think about that idea ?
Is there already a patch to do that, should I change rawlog.c or is ther already a regular way do do that ?
Thanks
-- Thomas Hummel | Institut Pasteur <hummel@pasteur.fr> | Pôle informatique - systèmes et réseau
On 3/08/2006 12:16 a.m., Thomas Hummel wrote:
Hello,
In rc2 sources, the rawlog directory seem to be hard-coded to be looked for in $HOME.
I'd like to have a way to centralized the place where rawlog files are, in order to be able :
. to rotate them more easily . not to leave them at the user's reach . not to have to enter the user's directory to read them
I'd like to have a layout similar to this :
/var/log/dovecot.rawlog/ /user1/20060708-093811-45581.in /user1/20060708-093811-45581.ou /user1/... /user2/20060708-093811-45590.in /user2/20060708-093811-45590.out
What do you think about that idea ?
These files are really only useful for debugging and tracing purposes, and Dovecot does not need them to operate. Do you need these files for a particular purpose?
Quite likely it'll just send your disk IO load through the roof, and degrade the performance of everyone trying to use your server. In other words, maybe not something you'd want turned on on a production server or one which you want reasonable performance out of. To have to deal with rotation suggests you want to leave this feature on for days on end, if not permanently.
OTOH what *may* be handy for debugging (and maybe for you) is the ability to selectively run rawlog for a specific username(s). That way it wouldn't matter much where the logs are kept as there would only be a single place (the specified users' home directory) where logs are actually generated.
Ditto with gdbhelper, which I think would be even more useful by being able to specify where the core files are dumped (although to be fair I haven't had a single crash or corefile from dovecot for months now). I changed the path in mine by hand but that's a bit of a hack.
Probably a bit late for 1.0 final, but would be potentially useful going forward............
Reuben
On Thu, Aug 03, 2006 at 12:33:32AM +1200, Reuben Farrelly wrote:
On 3/08/2006 12:16 a.m., Thomas Hummel wrote:
In rc2 sources, the rawlog directory seem to be hard-coded to be looked for in $HOME.
I'd like to have a way to centralized the place where rawlog files are, in order to be able :
These files are really only useful for debugging and tracing purposes, and Dovecot does not need them to operate. Do you need these files for a particular purpose?
Thanks for your answer. I was actually thinking of rawlog because of a mysterious case of message loss I had once on a previous beta or rc release (with thunderbird).
My idea was that, with rawlog on, you may have a proof that, in case of a message apparently "lost", it's not dovecot that loses it but the UA that ask for it's deletion.
Quite likely it'll just send your disk IO load through the roof, and degrade the performance of everyone trying to use your server. In other words, maybe not something you'd want turned on on a production server or one which you want reasonable performance out of.
I have to agree indeed. So maybe the idea described above is relevant only for a pre-production test.
-- Thomas Hummel | Institut Pasteur <hummel@pasteur.fr> | Pôle informatique - systèmes et réseau
On Thu, 2006-08-03 at 00:33 +1200, Reuben Farrelly wrote:
OTOH what *may* be handy for debugging (and maybe for you) is the ability to selectively run rawlog for a specific username(s). That way it wouldn't matter much where the logs are kept as there would only be a single place (the specified users' home directory) where logs are actually generated.
That's more or less how it works. If the user doesn't have ~/dovecot.rawlog/ directory, rawlog just execs imap process.
As for the special rawlog directory, sounds a bit like bloat. :)
participants (3)
-
Reuben Farrelly
-
Thomas Hummel
-
Timo Sirainen